An Experimental Study for the Effects of Noise on Pattern Recognition Algorithms.

Saved in:

Bibliographic Details
Title:	An Experimental Study for the Effects of Noise on Pattern Recognition Algorithms.
Authors:	Guang Yi Chen¹ guang_c@cse.concordia.ca, Krzyzak, Adam² krzyzak@cse.concordia.ca
Source:	Engineering Letters. May2025, Vol. 33 Issue 5, p1185-1192. 8p.
Subjects:	Pattern recognition systems, Additive white Gaussian noise, Fast Fourier transforms, Discrete wavelet transforms, Central processing units
Abstract:	Pattern recognition is a very important topic in computer vision. Among existing methods, which one is the most robust to noise? This is a very interesting question to answer. In this paper, we compare fifteen different methods for pattern recognition under different noise levels and different rotation angles. Most of these methods are invariant to translation, rotation, and scaling of the pattern images. Our experiments demonstrate that the Ridgelet + FFT (fast Fourier transform) descriptor is the most robust to additive Gaussian white noise (AGWN) for both a printed Chinese character dataset and an aircraft dataset. In addition, the Zernike moments and the Radon transform + FFT2 descriptor are also relatively robust to noise, but they are not as good as the Ridgelet + FFT descriptor for pattern recognition. We also compare the CPU (central processing units) computational time for all these methods for both pattern datasets. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Pattern recognition is a very important topic in computer vision. Among existing methods, which one is the most robust to noise? This is a very interesting question to answer. In this paper, we compare fifteen different methods for pattern recognition under different noise levels and different rotation angles. Most of these methods are invariant to translation, rotation, and scaling of the pattern images. Our experiments demonstrate that the Ridgelet + FFT (fast Fourier transform) descriptor is the most robust to additive Gaussian white noise (AGWN) for both a printed Chinese character dataset and an aircraft dataset. In addition, the Zernike moments and the Radon transform + FFT2 descriptor are also relatively robust to noise, but they are not as good as the Ridgelet + FFT descriptor for pattern recognition. We also compare the CPU (central processing units) computational time for all these methods for both pattern datasets. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X